To investigate the ability of patients with COPD to reproduce an exercise intensity accurately on the treadmill using dyspnea ratings obtained during incremental exercise on the cycle ergometer (cross-modal exercise prescription). Five visits over an 8-week period. Thirteen symptomatic patients with stable COPD. Age was 67+/-6 years (mean+/-SD). FEV1 was 1.15+/-0.22 L (45+/-7% predicted). At each visit, patients performed spirometry and exercise. Visit 1 was a practice incremental exercise test on the cycle ergometer. At visit 2 (1 week later), patients estimated the intensity of dyspnea using the 0 to 10 category-ratio scale during an incremental exercise test on the cycle ergometer (cycle estimation trial). Visit 3, 5 weeks later, was a practice session on the treadmill. At visit 4, 1 week later, patients were instructed to produce specific intensities of dyspnea (ie, dyspnea targets) at 50% and at anaerobic threshold (AT) or 80% of peak oxygen consumption (VO2) as calculated from results at visit 2 (treadmill production trial). Visit 5, 1 week later, was the treadmill estimation trial. Lung function was stable at all visits. Dyspnea ratings were 1.9+/-0.9 at 50% of VO2 and 5.6+/-1.5 at AT/80% of peak VO2 (17.5+/-3.3 mL/kg/min). The VO2 at the treadmill production trial (761+/-185 mL/min) was significantly higher than at the cycle estimation trial (612+/-159 mL/min) at the low dyspnea target (p < 0.0002; upward bias, 26+/-16%). In contrast, there was no significant difference in VO2 values (929+/-176 mL/min vs 948+/-259 mL/min) at the high dyspnea target (p > 0.5; 0+/-11% bias). Patients with COPD can use dyspnea ratings from an incremental cycle ergometry test to regulate exercise on the treadmill without systematic bias at an intensity of 80% of peak VO2, but exceed the desired VO2 when using the dyspnea rating at an intensity of 50% of peak VO2.